Pater Familias

WE KNOW LESS THAN COMPUTER MODELS SAY WE DO. My post earlier this week on findings by John Ioannidis that a sizable percentage of published scientific studies are later invalidated is important for understanding the current financial crisis. Studies based on computer models have a higher rate of invalidation than other studies. And people tend to overrate the accuracy of computer models. Any computer model is necessarily simplified. If it includes 500 variables, it still leaves all the others out. For example, the Long-Term Capital Management models in 1998 apparently assumed that liquid markets would continue to exist.

JOURNALS OF NEGATIVE RESULTS. The Economist article on the explanations advanced by Ioannidis for why so many scientific articles in prestigious journals are wrong says that “There also seems to be a bias towards publishing positive results…. even though negative results are potentially just as informative as positive results, if not as exciting.” For the last thirty years my brother Elmer has been saying that there should be a journal of negative results. There is hope. Now there are several journals of negative results, including journals in biomedicine, ecology and evolutionary biology, and natural language processing.

WHY ARE ALMOST ONE THIRD OF ARTICLES IN SCIENTIFIC JOURNALS WRONG? This article in the Economist for October 11 reports on an article by John Ioannidis and others which advances explanations for why so many published scientific articles turn out to be wrong. Dr. Ioannidis did a study three years ago of 49 papers in leading journals that had each been cited by more than 1000 other scientists. Within a few years almost one third of the papers had been refuted by later studies. Why does this happen? I posted previously on an article in the Economist subtitled “Why so much medical research is rot”, which pointed out that many articles are based on a misunderstanding of statistics. (Roughly, if I test 1000 hypotheses, on average I will get fifty results that I can claim are statistically significant–that I can say were “less than 5% likely to have come about by chance.”) Among the other explanations that Dr. Ioannidis and his colleagues advance are a bias toward publishing positive results and the “Winner’s Curse.”

EVOLUTIONARY PSYCHOLOGY AND NARRATIVE. If an ability to understand the thoughts of an imaginary character is present in a five year old, did that ability in humans arise by accident or because it provides an evolutionary advantage? In the article that I linked to in today’s other post (summary only), Jeremy Hsu discusses a very interesting article by Steven Pinker that addresses the question of whether there is an evolutionary advantage in being able to imagine other minds and to create stories. Pinker, as a scientist, raises objections: “One might have expected natural selection to have weeded out any inclination to engage in imaginary worlds rather than the real one.” I kept expecting Pinker to make an argument similar to that of David Foster Wallace– that it is an evolutionary advantage to be able to understand another human’s point of view– but I don’t think he does. (As opposed to Wallace, Pinker also questions the usefulness of studying literature rather than “patently useful subjects like biology or statistics.”) Instead, Pinker makes a case for one kind of evolutionary advantage from fiction: that it provides thought experiments which humans can use for “case-based reasoning.”

EMPATHY AND NARRATIVE.As I posted yesterday, David Foster Wallace argued that a function of a liberal arts education is to improve our ability to enter into the thoughts of others.
An article by Jeremy Hsu in the September Scientific American (summary only at the link) reports on studies of narrative by scientists. One of the studies claims that at the age of five children are able to follow the (highly simplified) thoughts of an imaginary character. Three year olds can’t do it. Wallace believes that it takes years of study of the liberal arts to perfect the raw ability of the five year old.

CLEAN CLOTHES FOR DOCTORS. The article that I posted on yesterday on whether hospital personnel should wear clean clothes says that “American hospitals operate on tight budgets and can’t afford to provide clothes and shoes to every worker. In addition, many hospitals don’t have the extra space for laundry facilities.” I don’t understand this resistance to trying to provide a sterile environment. Clean clothes are routine in nuclear facilities and places where computer chips are made. The importance of eliminating germs in hospitals has been known for over a hundred years, since the work of Semmelweis, Lister and Oliver Wendell Holmes (the father). (Holmes, the Autocrat of the Breakfast Table, is a model for this blog.)

CLEAN SURFACES (REVISITED). I have posted several times, including here, about efforts to reduce hospital infections. This article by Tara Parker-Pope describes how the British National Health Service has imposed a “bare below the elbows” rule against doctors wearing neckties and long sleeves. As reported here, a commentary in a British medical journal has questioned the rule because there is a lack of data. Ironically, a quote from the commentary says that the available evidence is “mostly in obscure medical journals.” I think that questions of hospital cleanliness don’t have a great deal of intellectual interest, so that it is not surprising that only obscure journals consider them. I also think that the controversy is a good example of the gulf between frequentists and Bayesians that I posted on here. There is evidence that germs can live a long time on fabrics. A Bayesian would give that fact, plus the fact that germs cause infections, a great deal of weight. But there is apparently no controlled experiment on the issue. As the Parker-Pope article says, “there’s no evidence that clothing plays a role in the spread of hospital infections.”

SCOTT FITZGERALD, CRITIC. I have been reading Bryant Mangum’s edition of THE BEST EARLY STORIES OF F. SCOTT FITZGERALD. Fitzgerald had remarkable range very early. It’s startling to find such wisdom in a young man who went on to lead such a troubled life. From the beginning, Fitzgerald combined lyricism with analytical distance. The Biographical Note contains a brilliant sentence which summarizes Fitzgerald’s writing as well as his life: “It was as if all his fiction described a big dance to which he had taken, as he once wrote, the prettiest girl…and as if he stood at the same time outside the ballroom, a little Midwestern boy with his nose to the glass, wondering how much the tickets cost and who paid for the music.” The quotation is from Malcolm Cowley, but you will note that Cowley is quoting Scott Fitzgerald.

DUELS—JAY AND CROW. Jack Sanders in our local Darien Times described a recent report of a duel between a Steller’s Jay and a crow.The jay tried without success to frighten a crow away from a feeder. The jay then broke a four-inch twig from a tree, put in his beak with the sharp end pointing out and flew at the crow. The crow got the twig away from the jay and wound up with the twig in its mouth chasing the jay. If dropping an object is not counted, this is apparently the first time a bird has been sighted using an object as a weapon. A biologist commented, “It’s a pity that the crow did not have its own stick to duel with the jay.”

CHOOSING AN ONCOLOGIST. Kids, this philosophical debate about the foundations of statistics is important. You may have noticed that I am a Bayesian. When we had to choose an oncologist, one of the factors turned out to be how they thought about statistical evidence (of course, brilliance, kindness and humanity were the most important factors). In that crucial first meeting, Dr. Rosenberg, the wonderful oncologist your mother chose, related statistics to his medical knowledge in the way a Bayesian would. The other oncologist that we talked to seemed to be reciting numbers out of context.